This paper describes the generation of Korean embedded sentences from the semantic structures in the interlingual machine translation systems. The semantic structure is internal representation which has not surface sentence information, so case-frame and Korean syntactic information is needed to generate surface sentences.
Korean sentences are composed of simple and complex sentences. The complex sentences are divided into conjunctive and embedded sentences. The embedded sentences are classified into six groups according to the criteria which generate embedded ending.
The generation of sentence is composed of two phases. First phase is syntactic generation which constructs the syntactic structure by traversing the semantic structure. Second phase is morphological generation which checks connectivity and deals with irregular conjugation. Since the generation is independent to the analysis, the results of this study can be adaptable to Korean natural language processing systems.
